Patterns and determinants of mammal species occurrence in India

Transcription

1 Journal of Applied Ecology 2009, 46, doi: /j x Patterns and determinants of mammal species occurrence in India Krithi K. Karanth 1,2 *, James D. Nichols 3, James E. Hines 3, K. Ullas Karanth 4 and Norman L. Christensen 1 1 Nicholas School of Environment, PO Box 90328, Duke University, Durham NC 27708, USA; 2 Department of Ecology, Evolution and Environmental Biology, Columbia University, NY 10027, USA; 3 U.S. Geological Survey Biological Resources Division, Patuxent Wildlife Research Center, Beech Forest Road, STE 4037, Laurel, MD , USA; and 4 Wildlife Conservation Society India Program, 403 Seebo Apts, 26-2 Aga Abbas Ali Road, Bangalore , India Summary 1. Many Indian mammals face range contraction and extinction, but assessments of their population status are hindered by the lack of reliable distribution data and range maps. 2. We estimated the current geographical ranges of 20 species of large mammals by applying occupancy models to data from country-wide expert. We modelled species in relation to ecological and social covariates (protected areas, landscape characteristics and human influences) based on apriorihypotheses about plausible determinants of mammalian distribution patterns. 3. We demonstrated that failure to incorporate detection probability in distribution survey methods underestimated habitat occupancy for all species. 4. Protected areas were important for the distribution of species. However, for many species much of their current range remains unprotected. The availability of evergreen forests was important for the occurrence of 14 species, temperate forests for six species, deciduous forests for 15 species and higher altitude habitats for two species. Low human population density was critical for the occurrence of five species, while culturally based tolerance was important for the occurrence of nine other species. 5. Rhino Rhinoceros unicornis, gaur Bos gaurus and elephant Elephas maximus showed the most restricted ranges among herbivores, and sun bear Helarctos malayanus, brown bear Ursus arctos and tiger Panthera tigris were most restricted among carnivores. While cultural tolerance has helped the survival of some mammals, legal protection has been critically associated with occurrence of most species. 6. Synthesis and applications. Extent of range is an important determinant of species conservation status. Understanding the relationship of species occurrence with ecological and socio-cultural covariates is important for identification and management of key conservation areas. The combination of occupancy models with field data from country-wide experts enables reliable estimation of species range and habitat associations for conservation at regional scales. Key-words: detection, distribution, India, land cover, mammals, occupancy, parks, people, range, spatial modelling Introduction *Correspondence author. krithi.karanth@gmail.com Many species now face large-scale range contraction and extinction (Channell & Lomolino 2000a,b; Laliberte & Ripple 2004; Cardillo et al. 2005). Human activities have locally extirpated animals and fragmented their habitats, thus influencing current species distribution patterns (Woodroffe & Ginsberg 1998; Sanderson et al. 2003). Determining where species occur and which species are threatened is a fundamental step in their conservation. Extent of range is widely recognized as an important criterion used to classify species according to conservation status. Examining the importance of nature reserves, different Ó 2009 The Authors. Journal compilation Ó 2009 British Ecological Society

2 1190 K. K. Karanth et al. land-cover land-use types, human densities and human cultural tolerance to species distribution and occurrence is critical to improving mammal conservation efforts (Cardillo et al. 2004). Large mammals are often keystone species that maintain ecosystem stability and biodiversity (Terborgh 1988). They are particularly vulnerable to local extirpations leading to drastic range contractions or even extinctions. Geographical endemism and inherent rarity exacerbated by human impacts have now pushed 25% of terrestrial mammal species close to extinction (Ceballos et al. 2005). Many species have disappeared from over 50% of their historical ranges (Ceballos & Ehrlich 2002). Large body sizes, unique habitat requirements, and associated life-history traits increase the vulnerability of mammals to extinction (Brashares 2003; Michalski & Peres 2005). Assessments of distributions of mammal species at the regional or subcontinental scale are difficult to obtain from direct field surveys of animals. This lack of basic information on species distributions significantly hinders mammal conservation (Brashares, Arcese & Sam 2001; Parks & Harcourt 2002). Most surveys (Channell & Lomolino 2000b; Sanderson et al. 2003; Ceballos et al. 2005) use questionnaires, static range maps or other forms of expert opinion as basic data. We note that such surveys are commonly affected by the problem of species being present at some locations but going unreported. As a result estimates of habitat occupancy are biased low, and the relationship between animal presence and habitat covariates remains poorly understood (MacKenzie et al. 2003; Tyre et al. 2003; Gu & Swihart 2004). India harbours 450 mammal species, has been occupied by humans for over years (Wells 2002), but also has a modern conservation history of regulating land uses to protect natural areas for over a century (Blythe 1863; Prater 1948; Rangarajan 2001). A country-wide wildlife reserve system was set up in the 1970s. In the last 100 years, rapid economic and human demographical growth has intensified human impacts on mammal species and their habitats (Forest Survey of India 2000; Das et al. 2006). Conservative estimates suggest that 20% of Indian large mammals face extinction, and several have disappeared from >90% of their original range (Madhusudan & Mishra 2003). The need for basic information on current mammal distributions and on ecological and social determinants of their persistence is acute. We conducted a country-wide survey based on field work of local experts and analysed resulting data using occupancy modelling to estimate the present geographical ranges of 20 large mammal species in India. We modelled species occurrence in relation to associated ecological and social covariates to elucidate observed species distribution patterns. These covariates (protected areas, landscape characteristics Table 1. Environmental and Social covariates and a priori predictions about their influence on habitat occupancy of large mammals in India Covariate category Variables Predictions for occupancy 1. Protected areas (pa, park) 2. Land cover land use (fc, lc) (a) Presence or absence of protected area (b) Proportion of cell covered by a protected area (5 categories: 0, 1 25%, 26 50%, 51 75% and %) (a) Presence or absence of forest cover (b) Ten categories derived from 23 original categories. Categories were evergreen forests, deciduous forests, temperate forests, barren, salt pan, scrub, cultivated, snow, water and rural-urban areas. Actual covariate within each category was the number of pixels of the specified habitat within the cell 3. Elevation (elv) Average elevation in a cell. Transformed by dividing by maximum elevation and multiplying by 10. Range of values between 0 and Human population Total number of people for every cell. Log density (ppl) transformation used 5. Cultural intolerance (tol) Ranked states from most to least tolerant states (1 to 3). Two states were ranked as most tolerant, nine as least tolerant, and the remaining states as medium tolerance Presence of protected areas and forest cover favours brown bear, black bear, gaur, elephant, rhino, Asiatic wild dog, tiger, leopard, sambar, muntjac, chital and sloth bear Evergreen forests favour sun bear, black bear, gaur, Asiatic wild dog, tiger, leopard, sambar, muntjac, chital, sloth bear and sun bear Deciduous forests favour all species Temperate forests favour brown bear and black bear Scrub-grassland favours all species except brown, black bear, tiger Barren and Salt Pan favour wolf, hyena, nilgai, blackbuck and chinkara Cultivated areas favour most herbivores and jackal Higher elevations favour brown bear, black bear and sun bear Low and Mid elevations favour gaur, wild pig, jackal, wolf, Asiatic wild dog, tiger, leopard, sambar, muntjac, chital, nilgai, blackbuck, chinkara, sloth bear and hyena Lowers occupancy for most species except wild pig and jackal Higher for herbivores (especially black buck, chinkara and nilgai) and lower for carnivores and wild pig Covariates: pa, park presence absence; park, percentage of cell covered; fc, forest cover presence absence; elv, elevation (transformed); ppl, log (human population density); tol, cultural intolerance. Land-cover and land-use categories are evergreen, temperate, deciduous, scrub grassland, barren, cultivated, rural urban, snow, salt pan and water.

3 Mammal occurrence and distribution in India 1191 Table 2. Best Models for predicting occupancy for each species Species Top-ranked model(s) with lowest AIC wi Model AIC n DAIC Chital Cervus axis w (e + t + s + d + park + tol) p (e + t + s + d + park + tol) 2 w (e + t + s + d + park + tol + ppl) p (e + t + s + d + park + tol + ppl) 2 w (e + s + d + c + b + park + tol + ppl) p (e + s + d + c + b + park + tol + ppl) 1 w (e + s + d + c + park + tol + ppl) p (e + s + d + c + park + tol + ppl) 1 Sambar Cervus unicolor w (e + s + d + c + park + ppl + tol) p (e + s + d + c + park + ppl + tol) 1 0Æ76 w (e + s + d + c + b + park + ppl + tol) p (e + s + d + c + b + park + ppl + tol) 1 0Æ21 Muntjac Muntiacus muntjak w (e + t + s + d + park + ppl + tol) p (e + t + s + d + park + ppl + tol) 2 0Æ88 w (e + s + d + park + ppl + tol) p (e + s + d + park + ppl + tol) 1 0Æ06 Blackbuck Antilope cervicapra w (e + s + sp + d + c + b + park + tol) p(e + s + sp + d + c + b + park + tol) 1 0Æ67 w (e + s + d + c + b + park + tol + ppl) p(e + s + d + c + b + park + tol + ppl) 1 0Æ18 w (s + d + c + b + tol) p (s + d + c + b + tol) 1 0Æ07 Nilgai Boselaphus tragocamelus w (s + d + c + b + park + tol) p (s + d + c + b + park + tol) 2 0Æ80 w (s + d + c + b + tol) p (s + d + c + b + tol) 2 0Æ20 Chinkara Gazella bennetti w (e + s + d + c + b + park + ppl + tol) p (e + s + d + c + b + park + ppl + tol) 1 0Æ Æ26 18 Wild pig Sus scrofa w (e + t + s + d + b + c + sp + sn + u + park + tol) p (e + t + s + d + b + c + sp + sn + u + park + tol) 3 w (e + t + s + d + b + c + sp + sn + u + park + tol + ppl) p (e + t + s + d + b + c + sp + sn + u + park + tol + ppl) 3 0Æ37 0Æ24 0Æ17 0Æ12 0Æ56 0Æ Æ Æ Æ Æ Æ Æ Æ Æ Æ Æ Æ Æ Æ Æ Æ Æ86 1Æ50 2Æ19 2Æ54 5Æ30 2Æ58 4Æ60 2Æ77 0Æ49 Gaur Bos gaurus w (e + s + d + park + ppl + tol) p (e + s + d + park + ppl + tol) 1 0Æ59 w (e + s + d + c + park + ppl + tol) p (e + s + d + c + park + ppl + tol) 1 0Æ15 w (e + t + s + d + park + ppl + tol) p (e + t + s + d + park + ppl + tol) 2 0Æ11 w (e + s + d + c + b + park + ppl + tol) p (e + s + d + c + b + park + ppl + tol) 1 0Æ10 Elephant Elephas maximus w (e + s + d + c + b + park + ppl + tol) p (e + s + d + c + b + park + ppl + tol) 1 0Æ87 w (e + s + d + park + ppl + tol) p (e + s + d + park + ppl + tol) 1 0Æ04 Rhino Rhinoceros unicornis w (elv + park) p (elv + park) 3 0Æ98 172Æ98 6 Black bear Ursus thibetanus w (e + t + s + d + tol + park) p (e + t + s + d + tol + park) 2 0Æ Æ54 14 w (e + t + s + d + tol + park + ppl) p (e + t + s + d + tol + park + ppl) 2 0Æ Æ29 6Æ75 Brown bear Ursus arctos w (e + t + s + d + park) p (e + t + s + d + park) 2 0Æ81 602Æ24 12 w (e + t + s + d + park + ppl) p (e + t + s + d + park + ppl) 2 0Æ12 606Æ Æ85 Sloth bear Melursus ursinus w (e + s + d + c + b + park + ppl) p (e + s + d + c + b + park + ppl) 1 1Æ Æ15 18 Sun Bear Helarctos malayanu) w (elv + park) p (elv + park) 3 6 w (park) p (.) 3 w (park) p (.) 3 w (park) p (park) 3 w (elv + park) p (.) 3 w (t) p (.) 1 w (.) p (t) 1 w (.) p (e + t + s + d) 2 w (elv) p (elv) 3 w (.) p (e) 1 Jackal Canis aureus w (e + t + s + d + b + c + sp + sn + u + park + ppl + tol) p (e + t + s + d + b + c + sp + sn + u + park + ppl + tol) 3 w (e + t + s + d + b + c + sp + park + ppl + tol) p (e + t + s + d + b + c + sp + park + ppl + tol) 3 0Æ21 0Æ08 0Æ07 0Æ07 0Æ04 0Æ03 0Æ03 0Æ03 0Æ02 0Æ02 0Æ64 0Æ Æ Æ Æ Æ Æ Æ08 176Æ54 178Æ50 178Æ58 178Æ71 179Æ63 180Æ18 180Æ25 180Æ44 180Æ90 181Æ Æ Æ Æ70 3Æ32 3Æ48 5Æ91 1Æ96 2Æ04 2Æ17 3Æ09 3Æ64 3Æ71 3Æ90 4Æ36 4Æ57 1Æ20

4 1192 K. K. Karanth et al. Table 2. (Continued) Species Top-ranked model(s) with lowest AIC w i Model AIC n DAIC Hyena Hyaena hyaena w (e + s + d + c + b + ppl + park + tol) p(e + s + d + c + b + ppl + park + tol) 2 1Æ Æ94 18 Wolf Canis lupus w (e + t + s + sp + d + c + b + ppl) p (e + t + s + sp + d + c + b + ppl) 2 1Æ Æ22 18 Asiatic wild dog Cuon alpinus w (e + t + s + d + tol + park) p (e + t + s + d + tol + park) 2 3Æ Æ Æ17 0Æ82 w (e + s + d + c + tol + park + ppl) p (e + s + d + c + tol + park + ppl) 1 0Æ15 Tiger Panthera tigris w (e + s + d + c + b + park + ppl + tol) p (e + s + d + c + b + park + ppl + tol) 1 0Æ Æ22 18 Leopard Panthera pardus w (e + t + s + d + b + c + sp + sn + u + park + ppl + tol + elv) p (e + t + s + d + b + c + sp + sn + u + park + ppl + tol + elv) 3 0Æ Æ65 28 Top-ranked models are shown, w i is the AIC model weight, n is the number of parameters in the model and DAIC is the difference in values between lowest AIC model and each model. All models are shown in Supporting Information Appendices S1 and S2. e, evergreen; t, temperate; s, scrub grassland; d, deciduous; c, cultivated; b, barren; sn, snow; u, urban rural; sp, salt pan; park, park proportion; ppl, human population density; tol, cultural intolerance; elv, elevation. 1 Individual species models. 2 Group-based models. 3 General models. and human influences) allowed us to test hypotheses about plausible determinants of species distribution patterns. We expected the presence of protected areas to have a positive effect and human population density to have a negative effect on the persistence of many species. We expected human cultural tolerance to show subregional variations and have different effects on herbivores and carnivores. The expected effects of landscape covariates varied among species. We developed species-specific occupancy models based on these hypotheses to test our predictions, and we developed corresponding distribution maps. Materials and methods OCCUPANCY MODELLING, EXPERT FIELD SURVEYS AND STUDY DESIGN Inferences about species distribution are complicated by locations that are not surveyed and by non-detection of species that are present in locations that are surveyed. Although the problem of non-detection and false absences occurrence has been recognized for some time (e.g. Preston 1948; Connor & McCoy 1979; Rosenzweig 1995), satisfactory approaches for its solution have only been developed relatively recently. Approaches for inference about species distributions based on so-called presence-only data (locations of species detections) have been developed (e.g. Anderson 2003; Phillips, Anderson & Schapire 2006) but are susceptible to various problems associated with sampling and non-detection. Occupancy studies are designed to provide inferences about the presence of target organisms in geographical sample units in the presence of imperfect detection. Occupancy models use some form of replication to deal with false absences, situations where species may be present but not always detected by observers (detection probabilities <1, MacKenzie et al. 2002, 2006). We used a grid-based sampling approach and divided the country into 1326 grid cells (average cell size 2818 km 2 ). We chose this grid cell size as it was practical to get sufficient replication of reports by local experts on presence of multiple species, as well as other data on ecological and social covariates for the entire country only at this scale. For each mammal species, we used existing information (Prater 1948; Menon 2003) to identify subsets of cells for which species occurrence was biogeographically and ecologically plausible and excluded areas where the species was historically absent. As a result of the scale of our survey, we obtained species presence absence data from knowledgeable Indian wildlife experts based on their field observations rather than conduct our own field surveys. Survey forms were completed by these experts (>100) between January and August Experts were selected based on their knowledge of particular regions and species. Experts were explicitly instructed to indicate the presence of each species only if they had personally observed either the species or its direct signs (tracks or scats) in the field within the specific grid cell(s) of interest in the past year. They were instructed not to indicate presence of a species if there was any uncertainty (e.g. in identifying scats or tracks to species) associated with an observation. Therefore, we did not interpret reports of non-detection to mean absence of the species, but a reported detection was interpreted as reflecting presence of the species. By emphasizing that all reports of presence had to represent certainty, we sought to eliminate problems of false presences (e.g. Royle & Link 2006). The number of observers (experts) surveyed per cell ranged between two

5 Mammal occurrence and distribution in India 1193 and 37, providing the replicate detection non-detection data needed to estimate occupancy. We used program presence (v 2.0; Hines 2006) to model detection probabilities and estimate occupancy for all 20 species. ESTIMATION AND MODELLING OF OCCUPANCY AND DETECTION PROBABILITY We estimated occupancy using the single-season models developed by MacKenzie et al. (2002, 2006), which use the method of maximum likelihood. For each species, we fit multiple models, representing different hypotheses about the processes that generated the data. The models were ranked in order of parsimony, and model weights were calculated using Akaike s Information Criterion (AIC; Burnham & Anderson 2002, 2004). The AIC weights sum to 1 for all members in a model set, and represent relative measures of the appropriateness of a given model relative to other models in the model set. We used model averaging based on AIC model weights in situations where there were multiple models that were supported by the data. Our approach to inference yields unconditional estimates of occupancy, ^w i for each cell, i, within the area deemed plausible for species occurrence. These values reflect the probability that cell i is occupied by the focal species, based on the covariate values associated with the cell. These probabilities can be viewed as expectations associated with a stochastic process and do not represent realizations of this process. Thus, we are not conditioning on the cells with detections (by assigning them probabilities of occurrence of 1) and then estimating conditional occupancy probabilities only for cells at which the species was not detected (see discussions in MacKenzie et al. 2006, pp and ). The unconditional occupancy estimates were then summed over all cells to estimate the overall proportion of plausible land area occupied by the species: ^w ¼ð P s ^w i¼1 i Þ=s, where s is the number of cells designated as plausible. This value was used to estimate the overall proportion of land (expressed at the scale of cells) within India that is occupied by the species: ^w IN ¼ a^w, where a = the proportion of land in all of India that was assigned to be plausible for the species. This proportion, a, was estimated by dividing the number of cells used in our analysis for a species by the total number of cells (1326) in India. COVARIATES In most field situations, occupancy and detection probabilities are not constant across all sample units (Royle 2005) but instead vary by site characteristics. We model occupancy and detection probabilities as functions of covariates using logit link functions (MacKenzie et al. Table 3. Estimated beta coefficients for the top ranked models for herbivores Species covariates Chital Sambar Muntjac Blackbuck Nilgai Chinkara Wild pig Gaur Elephant Rhino Constant )1Æ45 4Æ80 )2Æ99 3Æ88 3Æ 9Æ07 3Æ15 )12Æ77 )22Æ53 13Æ04 Evergreen 1Æ17 1Æ95 1Æ75 0Æ02 NA )1Æ52 4Æ64 1Æ11 3Æ83 NA Temperate )0Æ14 NA )0Æ39 NA NA NA )0Æ22 NA NA NA Scrub )0Æ32 0Æ15 )0Æ78 )1Æ14 0Æ69 0Æ47 )0Æ37 0Æ50 0Æ11 NA Snow NA NA NA NA NA NA )3 NA NA NA Rural urban NA NA NA NA NA NA 0Æ25 NA NA NA Salt pan NA NA NA 0Æ12 NA NA 0Æ05 NA NA NA Deciduous 0Æ48 0Æ52 0Æ68 0Æ 0Æ47 0Æ54 0Æ03 0Æ41 )0Æ28 NA Cultivated NA 0Æ13 NA 0Æ29 0Æ20 NA 0Æ06 NA )0Æ41 NA Barren NA NA NA 0Æ46 )0Æ08 0Æ42 0Æ34 NA )0Æ72 NA Elevation NA NA NA NA NA NA NA NA NA )4Æ96 Park 0Æ74 0Æ61 0Æ50 0Æ12 0Æ13 0Æ49 0Æ22 0Æ06 )0Æ51 )12Æ78 Population density NA )0Æ58 )0Æ37 NA NA )0Æ53 NA 0Æ46 1Æ46 NA Tol (intolerance) 0Æ28 0Æ78 3Æ20 )2Æ91 )2Æ47 )3Æ37 )0Æ93 1Æ88 2Æ57 NA Table 4. Estimated beta coefficients for the top ranked models for carnivores Species covariates Black bear Brown bear Sloth bear Sun bear Jackal Hyena Wolf Asiatic wild dog Tiger Leopard Constant )5Æ85 )1Æ61 )7Æ93 )0Æ68 )0Æ14 2Æ31 0Æ59 )9Æ21 )10Æ33 12Æ74 Evergreen 0Æ35 )6Æ35 )0Æ21 NA 0Æ17 0Æ06 0Æ21 0Æ66 0Æ31 1Æ57 Temperate 15Æ87 4Æ70 NA NA 0Æ12 NA 1Æ10 0Æ01 NA 5Æ18 Scrub )0Æ28 )0Æ10 0Æ53 NA )0Æ03 )0Æ32 0Æ28 )0Æ02 )0Æ01 )0Æ18 Snow NA NA NA NA )0Æ27 NA NA NA NA )1Æ35 Rural urban NA NA NA NA 0Æ06 NA NA NA NA 0Æ57 Salt pan NA NA NA NA 0Æ04 NA 0Æ70 NA NA )0Æ40 Deciduous )0Æ63 )0Æ10 0Æ80 NA 0Æ15 0Æ61 0Æ76 0Æ38 0Æ41 0Æ45 Cultivated NA NA )0Æ29 NA 0Æ09 0Æ17 0Æ56 NA 0Æ10 0Æ13 Barren NA NA 0Æ75 NA 0Æ06 0Æ06 )0Æ22 NA )0Æ29 0Æ38 Elevation NA NA NA )1Æ08 NA NA NA NA NA )0Æ32 Park )0Æ71 1Æ68 0Æ50 2Æ86 0Æ32 0Æ36 NA 0Æ29 0Æ91 0Æ50 Population density NA NA 0Æ76 NA 0Æ39 )0Æ19 NA NA 0Æ42 )0Æ99 Tol (intolerance) 2Æ99 NA )0Æ69 NA )1Æ92 )0Æ60 )0Æ22 3Æ29 0Æ96 )0Æ23

6 1194 K. K. Karanth et al. 2006). For example, the logit of the probability of a site being occupied is expressed as Logit ðw i Þ¼b 0 þ b 1 x i1 þ b 2 x i2 þ :::::::: þ b u x iu ; which is a linear function of the u covariates associated with site i, with one intercept term b 0 and u regression coefficients that need to be estimated (MacKenzie et al. 2006). Based on ecological and social contextual knowledge, we selected covariates most likely to influence the distribution of large mammals in India. These include designated wildlife protected area in the cell, landscape characteristics and human influences. Data on protected areas are from the World Database on Protected Areas ( and were further refined using topographic maps and expert knowledge from India. The data were used to create two measures as covariates presence or absence of a protected area in a cell, and proportion of cell covered by a protected area (Table 1). We selected presence or absence of forest cover in a cell, land cover land use data, and elevation as covariates representing landscape characteristics. Land cover land use data were derived from Global Land Cover Facility 2000 (Bartholomé &Belward 2005) and refined using Joshi et al. (2006) and Roy et al. (2006). We consolidated land-cover and land-use categories from 23 to 10, and used the total number of pixels in each category for every grid cell. This allowed us to build a reasonable and easily interpretable model set (Table 1). Elevation data were obtained from CGIAR-CSI SRTM 100 m (CGIAR-CSI 2004) Digital Elevation Data (2004). We calculated the average elevation in every grid cell, divided by the maximum, and transformed this to range between 0 and 10. To measure human influence, we used human population density and human cultural tolerance towards mammal species that exist in India. Human population density data were derived from LandScan Global Population Data 2000 ( (Dobson et al. 2000; Budhendra et al. 2002). We calculated human population density for every cell, and log transformed this variable. We developed a human cultural tolerance variable from prior personal observations, socio-cultural knowledge and hunting patterns of local communities in the different states of India (Rangarajan 2001; Madhusudan & Karanth 2005; Datta 2007; K. K. Karanth and K. U. Karanth, pers. obs.). India is a multi-cultural country and the State boundaries were reorganized in the 1950s and 1960s in recognition of the fact that linguistic affinities represent overall diversity effectively. Furthermore, effectiveness of anti-hunting law enforcement that shapes the degree of hunting pressure is influenced by each State s administrative culture (based on factors like efficiency, corruption, management policies and quality of field personnel). For example, States such as Madhya Pradesh and Karnataka have a relatively strong culture of law enforcement, whereas the ethnically tribal hill states in north-eastern India have poor law enforcement and much higher levels of illegal hunting. Therefore, we believe this covariate effectively represents both local people s cultural tolerance and the official law enforcement effectiveness. We grouped States from most tolerant to least tolerant (Table 1). The western states of Rajasthan and Gujarat were classified as most tolerant, seven north-eastern hill states, Chhattisgarh and Jharkhand were classified as least tolerant, and all other states as medium tolerance. We fit several general, group and individual species-based occupancy models (Table 2). We grouped the species based on apriori knowledge of their general habitat preferences (Table 2). Group 1 had 14 species that are believed to prefer more closed habitat types (deciduous, evergreen, temperate and scrub-grassland areas). Group 2 had six species believed to prefer more open habitat types (scrub-grassland, salt pan, barren and cultivated areas). We also ran occupancy models tailored to reflect hypothesized habitat preferences of individual species. For rare species (e.g. sun bear Helarctos malayanus), our model set was smaller and restricted to simpler models. For each species, we estimated the total area (expressed as number of cells) currently occupied by summing occupancy across all cells (Table 5). Table 5. Estimated occupancy for large mammals in India Species Common name x s Naı ve estimate of w R^w i ^w ^win Cervus axis Chital Æ48 483Æ23 0Æ48 0Æ36 Cervus unicolor Sambar Æ48 721Æ82 0Æ63 0Æ53 Muntiacus muntjak Muntjac Æ39 505Æ41 0Æ47 0Æ39 Antilope cervicapra Blackbuck Æ35 588Æ02 0Æ61 0Æ44 Boselaphus tragocamelus Nilgai Æ57 557Æ03 0Æ63 0Æ42 Gazella bennetti Chinkara Æ44 430Æ95 0Æ49 0Æ32 Sus scrofa Wild pig Æ Æ78 0Æ91 0Æ84 Bos gaurus Gaur Æ21 252Æ04 0Æ31 0Æ19 Elephas maximus Elephant Æ20 340Æ36 0Æ35 0Æ26 Rhinoceros unicornis Rhino Æ08 75Æ62 0Æ46 0Æ06 Ursus thibetanus Black bear Æ43 201Æ62 0Æ67 0Æ15 Ursus arctos Brown bear Æ42 91Æ92 0Æ54 0Æ07 Melursus ursinus Sloth bear Æ42 762Æ37 0Æ68 0Æ57 Helarctos malayanus Sun bear Æ14 4 0Æ30 0Æ03 Canis aureus Jackal Æ Æ58 0Æ83 0Æ78 Hyaena hyaena Hyena Æ57 671Æ13 0Æ65 0Æ51 Canis lupus Wolf Æ52 894Æ79 0Æ82 0Æ67 Cuon alpinus Asiatic wild dog (Dhole) Æ19 330Æ99 0Æ30 0Æ25 Panthera tigris Tiger Æ21 323Æ81 0Æ27 0Æ24 Panthera pardus Leopard Æ52 904Æ21 0Æ73 0Æ68 The number of cells in which a species was detected = x and the number of plausible cells within which a species might occur = s. The naïve estimate of occupancy for plausible cells is w = x s. ^w i is the estimated occupancy probability for the ith cell and ^w = R^w i s is the average of the estimated probabilities of occupancy for plausible cells, computed as the sum of occupancy probabilities for all plausible cells divided by the number of cells. ^w IN is R^w i divided by 1326 (the total number of cells in India).

7 Mammal occurrence and distribution in India 1195 APRIORIPREDICTIONS Modelling covariates enabled us to evaluate predictions from apriori hypotheses about factors influencing probabilities of occupancy and detection. For protected area covariates, we predicted higher occupancy for 10 species that prefer dense habitat cover (Table 1). We selected land-cover covariates that are ecologically relevant (Table 1; Supporting information Appendices S1 and S2). We also predicted elevation to affect occupancy, with higher altitude predicted to yield increased occupancy for three bear species and reduced occupancy for other species (Table 1). We expected human population density to have a negative relationship with occupancy for most species (except wild pig Sus scrofa and jackal Canis aureus, which adapt to settlements, and nilgai Boselaphus tragocamelus, chinkaragazella bennetti, and blackbuck Antilope cervicapra that are culturally tolerated in some areas). For the cultural tolerance covariate, we expected greater tolerance for and thus higher occupancy for herbivores as opposed to carnivores. Models include effects of covariates on probabilities of occupancy, detection, and both occupancy and detection. Detailed predictions are shown in Table 1. Latin names are in Table 2. Results PROPORTION OF PROTECTED AREA The proportion of protected area was included in at least one of the top models for 19 species (exception for wolf Canis lupus; Table 2). The b parameters were positive for species (Tables 3 and 4), indicating that the proportion of land set aside as protected area is important to species occupancy. Our predictions were correct for 10 forest-dwelling species, but extended to other species as well. LAND COVER AND ELEVATION Different combinations of land-cover types appeared important for different species. Evergreen forest was in the top model for 17 species (all except nilgai, rhino Rhinoceros unicornis and sun bear Table 2). The b parameters for evergreen forest were positive for 14 of these species and negative for three (chinkara, brown bear Ursus arctos, sloth bear Melursus ursinus). Our predictions were supported for all species except sloth bear. Temperate forest was in the top model for nine species, b parameters were positive for six species, and were negative for the other three species (Tables 3 and 4).Our predictions were found true for these species and extended to jackal, wolf, hyena Hyaena hyaena and leopard Panthera pardus. Deciduous forest was in the top model of 18 species (exceptions were rhino and sun bear; Table 2). The b parameters for deciduous forests were positive for 15 species. Negative b parameters for elephant Elephas maximus, brown and black bear Ursus thibetanus, indicate that our predictions were correct for all except Blackbuck Nilgai Chinkara Chital N Probability of occurrence (psi) Fig. 1. Estimated probabilities of occurrence for blackbuck Antelope cervicapra, nilgai Boselaphus tragocamelus, chinkaragazella bennetti and chital Cervus axis.

8 1196 K. K. Karanth et al. these three species. For details of other land-cover and landuse types and their effects on individual species occurrence, see Tables 2 4. Elevation was included in the top models for three species (rhino, sun bear and leopard) with negative b parameters for all three. Our predictions were not confirmed for a positive relationship with elevation for three bear species brown bear, blackbearandsunbear. HUMAN POPULATION DENSITY AND CULTURAL TOLERANCE Human population density was included in the top model for 10 species, and in additional models receiving substantive support for seven species. The b parameters were negative for three herbivores (sambar Cervus unicolor, muntjac Muntiacus muntjak, chinkara), and two carnivores (hyena and leopard). This suggests that human population density affects occupancy of some species, with some species adapting better to people (jackal) or coming into less contact with people (gaur Bos gaurus, elephant, sloth bear and tiger Panthera tigris) than others. Humanculturaltolerancewasincludedinthetopmodels for 17 species. The b parameters were negative for nine species (reflecting increased probability of occurrence with increasing tolerance), and positive for the eight others (Tables 3 and 4). The species with negative b parameters are adaptable and generally tolerated species (blackbuck, nilgai, chinkara, wild pig, sloth bear, jackal, hyena, wolf and leopard). Species with positive b parameters are mainly restricted to protected areas or are feared by people. Our predictions were supported for most species. ESTIMATED AREA OCCUPIED BY DIFFERENT MAMMALS We summed occupancy estimates and developed predicted occurrence and distribution maps for all 20 species (Table 5; Figs 1 5). For all species, the estimated proportion of cells occupied by a species was higher than the naı ve occupancy, especially so for sambar, blackbuck, wild pig, gaur, elephant, rhino, black bear, brown bear, sloth bear, sun bear, wolf, Asiatic wild dog Cuon alpinus and leopard (Table 5; Figs 1 5). The most widespread species were jackal (total ^w IN =0Æ78; Fig. 5) and wild pig (total ^win =0Æ91; Fig. 2). The most restricted species were rhino (^w IN =0Æ06; Fig. 2), and sun bear (^w IN =0Æ03; Fig. 4). Among herbivores the ^w values (estimated probability that an average plausible cell is occupied) ranged from 0Æ31 to 0Æ91. The herbivores with the most restricted ranges are rhino, gaur and elephant (Table 5). For carnivores, ^w values ranged more widely from 0Æ27 to 0Æ83. The Wild pig One-horned rhino Elephant Gaur N Probability of occurrence (psi) Fig. 2. Estimated probabilities of occurrence for wild pig Sus scrofa, rhino Rhinoceros unicornis, elephant Elephas maximus and gaur Bos gaurus.

9 Mammal occurrence and distribution in India 1197 carnivores most restricted in range are sun bear, brown bear and tiger (Table 5). Discussion Range maps are widely used to understand species distribution, focus conservation efforts and classify species by conservation status. In India, current range maps for most large mammal species are unreliable. We used occupancy modelling to map species distributions and investigate determinants of species occurrence. Our results show that failure to incorporate detection probability in surveys and analyses substantially underestimates overall habitat occupancy of all species (Table 5). We recommend that distribution surveys based on animal presence versus absence data should be designed a priori to incorporate imperfect detections to overcome this bias. In future analyses of species range, we believe, it will be useful to investigate auto-logistic models that explicitly model focal cell occupancy as a function of occupancy of cells within a specified neighbourhood (e.g. Augustin, Mugglestone & Buckland 1996). These models have been recently used for inference about occupancy in the presence of non-detection (Sargeant et al. 2005; Royle & Dorazio 2008), and we expect them to see greater use as software becomes available to facilitate implementation. We found protected areas are key determinants of of species, particularly those which are restricted regionally, or require forested habitats. Many species (e.g. tiger, Asiatic wild dog and elephant) that had a wide distribution <100 years ago (K.K. Karanth, J.D. Nichols, U.K. Karanth, J.E. Hines & N.L. Christensen, unpublished), are now primarily restricted to protected areas (Figs 2, 3 and 5). Although India s protected areas are small (average size <300 km 2 ) and fragmented, their carrying capacities for large mammals are high (Karanth et al. 2004). Improving protection and monitoring animal population status will be critical to their survival. Given that protected areas occupy <4% of the country s land area now, expanding India s nature reserves, reducing their fragmentation and improving connectivity, as well as enhancing legal and traditional protection mechanisms, should be conservation policy priorities. Maintaining landscape diversity, connectivity and compatibility of mammal habitats with human land uses will be also important for species whose habitats are substantially outside protected reserves (e.g. blackbuck, chinkara, wild pig, hyena, jackal, sloth bear, wolf and leopard; Figs 1 5). Rapid ongoing land-use changes (driven by human demographic growth) and Samber Muntjac Tiger Leopard N Probability of occurrence (psi) Fig. 3. Estimated probabilities of occurrence for sambar Cervus unicolor, muntjac Muntiacus muntjak, tiger Panthera tigris and leopard Panthera pardus.

10 1198 K. K. Karanth et al. Sun bear Asiatic black bear Sloth bear Brown bear N Probability of occurrence (psi) Fig. 4. Estimated probabilities of occurrence for sun bear Helarctos malayanus, black bear Ursus thibetanus, sloth bear Melursus ursinus and brown bear Ursus arctos. economic development present significant challenges to protecting these species. Forest clearance for agriculture in historical times has actually benefited species such as blackbuck, chinkara, nilgai, wolf and jackal. Species that are likely to benefit from the predominant Hindu cultural tolerance include nilgai, blackbuck and chinkara (in parts of Northern and Western India), gaur (in central and southern India), and elephants, macaques and langurs across much of their range, exemplifying this phenomenon unique to the Indian subcontinent. Our results support previous studies (Newmark 1995, 1996; Woodroffe & Ginsberg 1998; Brashares et al. 2001; Parks & Harcourt 2002; Treves & Karanth 2003; Cardillo et al. 2004) showing human demography and land use are important factors in mammal distribution, but cultural tolerance is a unique additional positive factor in India despite high human densities. It is particularly noteworthy that all legal hunting has been virtually banned for over 30 years in India taking advantage of this cultural ethos. Therefore, policies such as game harvesting or trophy hunting appear to be inappropriate as conservation tools in India despite their potential utility elsewhere in the world. The ranges of several of the studied species of large mammals have contracted greatly. Some of these species will require expanded protected areas with strong law enforcement if they are to survive, while others will require additional strategies based on modifying human land uses based on traditional social tolerance or current non-consumptive utilitarian values. Our results are likely to be useful in crafting conservation strategies that appropriately blend protected areas, land-use changes, law enforcement and cultural traditions to ensure the persistence of India s large mammals at large regional scales. Acknowledgements The authors are enormously grateful to the wildlife experts N. Akhtar, R. Ali, S. Amu, A. Aiyadurai, Y. V. Bhatnagar, R. Borges, A. Chandola, S. Chandola, D. Chetry, K. Choudhary, S. Choudhry, H. Dang, S. Dasgupta, S. Dattatri, A. Dutta, P. S. Easa, D. V. Girish, H. Ghuleria, D. Ghose, S. P. Goyal, B. Hegde, Hilaluddin, D. Jathanna, B. Jetva, Y. V. Jhala, A. J. T. John Singh, S. Jones, J. Joshua, K. Kakati, K.U. Karanth, K. Kathju, D. K. Kashyap, R. Kaul, M. Khanduja, J. Kulkarni, A. Kumar, N. S. Kumar, S. Kumar, H. Kumara, A. Lobo, P. Mehta, B. Mohanty, S. Molur, S. Mukherjee, L. Nehemiah, N. Patil, S. Pawar, S. Pradhan, S. Radhakrishna, A. Rahmani, N. Rajamani, S. Rajesh, S. Ram, G. S. Rawat, G. V. Reddy, V. Rishi, A. D. Roy, P. K. Sen, K. Sharma, N. Sharma, K. Sathasivam, J. N. Shah, G. Shahabuddin, D. Sharma, V. Srinivas, G. Sundar, A. Tamim, S. Tiwari, P. Trivedi, H. Tyabji, N. Ved, R. Vyas, R Wangchuk, T. Wangyal, and their associates who completed the surveys. We thank S. L. Pimm, D. L. Urban, C.N. Jenkins, M. Vale, N. S. Kumar, P. M. Kumar, Duke University, Center for Wildlife Studies and Wildlife Conservation Society s India Program. Karanth received funding from the Conservation, Food and Health Foundation, Forest History Society, Duke International Travel, IDEA Wild, Roger Williams Park Zoo, AZFA Clark Waldram Conservation Fund, and Cleveland Zoo. We thank the editors and referees for their constructive comments that led to substantial improvements to the manuscript.

11 Mammal occurrence and distribution in India 1199 Jackal Dhole (Asiatic wild dog) Wolf Striped hyena N Probability of occurrence (psi) Fig. 5. Estimated probabilities of occurrence for jackal Canis aureus, Asiatic wild dog Cuon alpinus, wolf Canis lupus and hyena Hyaena hyaena. References Anderson, R.P. (2003) Real vs. artefactual absences in species distributions: test with Oryzomys albigularis (Rodentia: Muridae) in Venezuela. Journal of Biogeography, 30, Augustin, N.H., Mugglestone, M.A. & Buckland, S.T. (1996) An autologistic model for the spatial distribution of wildlife. Journal of Applied Ecology, 33, Bartholome, E. & Belward, A.S. (2005) GLC2000: a new approach to global land cover mapping from earth observation data. International Journal of Remote Sensing, 26, Blythe, E. (1863). Catalogue of the Mammalia in the Museum Asiatic Society. Savielle and Cranenburgh. Bengal Printing Committee Ltd, London, UK. Brashares, J.S. (2003) Correlates of extinction in West Africa. Conservation Biology, 17, Brashares, J.S., Arcese, P. & Sam, M.K. (2001) Human demography and reserve size predict wildlife extinction in West Africa. Proceedings of the Royal Society of London, 268, Budhendra, B., Bright, E., Coleman, P. & Dobson, J. (2002) LandScan: locating people is what matters. Geoinformatics, 5, Burnham, K.P. & Anderson, D.R. (2002) Model Selection and Multimodal Inference: A Practical Information Theoretic Approach. Springer-Verlag, New York, NY. Burnham, K.P. & Anderson, D.R. (2004) Multimodal inference: understanding AIC and BIC in model selection. Sociologial Methods and Research, 33, Cardillo, M., Purvis, A., Sechrest, W., Gittleman, J.L., Beilby, J. & Mace, G.M. (2004) Human population density and extinction risk in the world s carnivores. Public Library of Science Biology, 2, Cardillo, M., Mace, G.M., Jones, K.E., Bielby, J., Bininda-Emonds, O.R.P., Sechrest, W., Orme, C.D.L. & Purvis, A. (2005) Multiple causes of high extinction risk in large mammal species. Science, 309, Ceballos, G. & Ehrlich, P.R. (2002) Mammal population losses and the extinction crisis. Science, 296, Ceballos, G., Ehrlich, P.R., Soberon, J., Salazar, I. & Fay, J.P. (2005) Global mammal conservation: what must we manage? Science, 309, CGIAR-CSI (2004) SRTM data V1, International Centre for Tropical Agriculture (CIAT). (accessed June 2006). Channell, R. & Lomolino, M. (2000a) Trajectories to extinction: spatial dynamics of the contraction of geographic ranges. Journal of Biogeography, 27, Channell, R. & Lomolino, M. (2000b) Dynamic biogeography and the conservation of endangered species. Nature, 403, Connor, E.F. & McCoy, E.D. (1979) The statistics and biology of the speciesarea relationship. American Naturalist, 113, Das, A., Krishnaswamy, J., Bawa, K.S., Kiran, M.C., Srinivas, V., Kumar, N.S. & Karanth, U.K. (2006) Prioritization of conservation areas in the Western Ghats, India. Biological Conservation, 133, 31. Datta, A. (2007). Protecting with people in Namdapha: threatened forests, forgotten people. Making Conservation Work (eds G. Shahabuddin & M. Rangarajan), pp Permanent Black, New Delhi. Dobson, J.E., Bright, E.A., Coleman, P.R., Durfee, R.C. & Worley, B.A. (2000) LandScan: a global population database for estimating populations at risk. Photogrammetric Engineering and Remote Sensing, 66, Forest Survey of India (2000) State of the Forest Report. Forest Survey of India, Dehradun. Gu, W. & Swihart, R.K. (2004) Absent or undetected? Effects of non-detection of species occurrence on wildlife-habitat models Biological Conservation, 1, Hines, J.E. (2006). Program PRESENCE (Version 2). usgs.gov/software/doc/presence/presence.html. Joshi, P.K.K., Roy, P.S., Singh, S., Agrawal, S. & Yadav, D. (2006) Vegetation cover mapping in India using multi-temporal IRS Wide Field Sensor (WiFS) data. Remote Sensing of Environment, 103, Karanth, U.K., Nichols, J.D., Kumar, S.N., Link, W.A. & Hines, J.E. (2004) Tigers and their prey: predicting carnivore densities from prey abundance. Proceedings of the National Academy of Sciences, 101,

12 1200 K. K. Karanth et al. Laliberte, A.S. & Ripple, W.J. (2004) Range contractions of North American carnivores and ungulates. BioScience, 54, MacKenzie, D.I., Nichols, J.D., Lachman, G.B., Droege, S., Royle, J.A. & Langtimm, C.A. (2002) Estimating site occupancy rates when detection probabilities are less than one. Ecology, 83, MacKenzie, D.I., Nichols, J.D., Hines, J.E., Knutson, M.G. & Franklin, A.D. (2003) Estimating site occupancy, colonization and local extinction when a species is detected imperfectly. Ecology, 84, MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.H., Bailey, L.A. & Hines, J.E. (2006). Occupancy Modeling and Estimation. Academic Press, San Diego, CA. Madhusudan, M.D. & Karanth, U.K. (2005). Local hunting and large mammal conservation. Wildlife Conservation, Research and Management (eds Y.V. Jhala, R. Chellam & Q. Qureshi), pp Wildlife Institute of India, Dehradun. Madhusudan, M.D. & Mishra, C. (2003). Why big, fierce animals are threatened: conserving large mammals in densely populated landscapes. Battles Over Nature: Science and Politics of Conservation (eds V. Saberwal & M. Rangarajan), pp Permanent Black, New Delhi. Menon, V. (2003) A Field Guide to Indian Mammals. Dorling Kindersley India Pvt. Limited, Delhi. Michalski, F. & Peres, C.A. (2005) Anthropogenic determinants of primate and carnivore local extinctions in a fragmented forest landscape of southern Amazonia. Biological Conservation, 124, Newmark, W.D. (1995) Extinction of mammal populations in western North American national parks. Conservation Biology, 9, Newmark, W.D. (1996) Insularization of Tanzanian parks and the local extinction of mammals. Conservation Biology, 10, Parks, S.A. & Harcourt, A.H. (2002) Reserve size, local human density, and mammalian extinctions in U.S protected areas. Conservation Biology,, Phillips, S.J., Anderson, R.P. & Schapire, R.E. (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling, 190, Prater, S.H. (1948) The Book of Indian Animals. Bombay Natural History Society, Bombay. Preston, F.W. (1948) The commonness, and rarity, of species. Ecology, 29, Rangarajan, M. (2001) India s Wildlife History. Permanent Black, New Delhi. Rosenzweig, M.L. (1995) Species Diversity in Space and Time. Cambridge University Press, Cambridge. Roy, P.S., Joshi, P.K., Singh, S., Agrawal, S., Yadav, D. & Jegannathan, C. (2006) Biome mapping in India using vegetation type map derived using temporal satellite data and environmental parameters. Ecological Modelling, 1107, Royle, J.A. (2005) Site occupancy models with heterogeneous detection probabilities. Biometrics, 62, Royle, J.A. & Dorazio, R.M. (2008) Hierarchical Modeling and Inference in Ecology: The Analysis of Data From Populations. Acadmic Press, San Diego, CA. Royle, J.A. & Link, W.A. (2006) Generalized site occupancy models allowing for false positive and false negative errors. Ecology, 87, Sanderson, E.W., Jaiteh, M., Levy, M.A., Redford, K.H., Wannebo, A.V. & Woolmer, G. (2003) The human footprint and the last of the wild. BioScience, 52, Sargeant, G.A., Sovada, M.A., Slivinski, C.C. & Johnson, D.H. (2005) Markov chain Monte carlo estimation of species distributions: a case study of the swift fox in western Kansas. Journal of Wildlife Management, 69, Terborgh, J.W. (1988) The big things that run the world- a sequel to E.O Wilson. Conservation Biology, 2, Treves, A. & Karanth, U.K. (2003) Human-carnivore conflict and perspectives on carnivore management worldwide. Conservation Biology, 17, Tyre, A.J., Tenhumberg, B., Field, S.A., Niejalke, D., Parris, K. & Possingham, H.P. (2003) Improving precision and reduction bias in biological surveys, by estimating false negative error rates in presence-absence data. Ecological Applications, 13, Wells, S. (2002) The Journey of Man: A Genetic Odyssey. Princeton University Press, Princeton, NJ. Woodroffe, R. & Ginsberg, G.R. (1998) Edge effects and extinctions of populations inside protected areas. Science, 280, Received 19 March 2009; accepted 19 August 2009 Handling Editor: Andy Royle Supporting Information Additional Supporting Information may be found in the online version of this article. Appendix S1. Models included in the model sets for all species. Appendix S2 Models included in the model sets for groups of species. As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.